CN116417115B - Personalized nutrition scheme recommendation method and system for gestational diabetes patients - Google Patents
Personalized nutrition scheme recommendation method and system for gestational diabetes patients Download PDFInfo
- Publication number
- CN116417115B CN116417115B CN202310665199.9A CN202310665199A CN116417115B CN 116417115 B CN116417115 B CN 116417115B CN 202310665199 A CN202310665199 A CN 202310665199A CN 116417115 B CN116417115 B CN 116417115B
- Authority
- CN
- China
- Prior art keywords
- knowledge
- information
- nutrition
- recipe
- gestational diabetes
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 235000016709 nutrition Nutrition 0.000 title claims abstract description 150
- 230000035764 nutrition Effects 0.000 title claims abstract description 122
- 208000004104 gestational diabetes Diseases 0.000 title claims abstract description 100
- 238000000034 method Methods 0.000 title claims abstract description 49
- 235000005911 diet Nutrition 0.000 claims abstract description 68
- 230000037213 diet Effects 0.000 claims abstract description 67
- 238000000605 extraction Methods 0.000 claims description 58
- 230000004927 fusion Effects 0.000 claims description 36
- 239000011159 matrix material Substances 0.000 claims description 32
- 230000006870 function Effects 0.000 claims description 21
- 238000009826 distribution Methods 0.000 claims description 12
- 235000020785 dietary preference Nutrition 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 5
- 238000003062 neural network model Methods 0.000 claims description 5
- 238000013145 classification model Methods 0.000 claims description 4
- 238000003860 storage Methods 0.000 claims description 4
- 239000000284 extract Substances 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000035935 pregnancy Effects 0.000 claims description 3
- 235000013305 food Nutrition 0.000 description 14
- 230000008569 process Effects 0.000 description 13
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 8
- 201000010099 disease Diseases 0.000 description 7
- 235000015097 nutrients Nutrition 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 208000024891 symptom Diseases 0.000 description 5
- 239000008280 blood Substances 0.000 description 4
- 210000004369 blood Anatomy 0.000 description 4
- 230000036541 health Effects 0.000 description 4
- 206010012601 diabetes mellitus Diseases 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 238000002372 labelling Methods 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 235000014633 carbohydrates Nutrition 0.000 description 2
- 150000001720 carbohydrates Chemical class 0.000 description 2
- 235000020979 dietary recommendations Nutrition 0.000 description 2
- OVBPIULPVIDEAO-LBPRGKRZSA-N folic acid Chemical compound C=1N=C2NC(N)=NC(=O)C2=NC=1CNC1=CC=C(C(=O)N[C@@H](CCC(O)=O)C(O)=O)C=C1 OVBPIULPVIDEAO-LBPRGKRZSA-N 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 206010067484 Adverse reaction Diseases 0.000 description 1
- 206010018429 Glucose tolerance impaired Diseases 0.000 description 1
- OVBPIULPVIDEAO-UHFFFAOYSA-N N-Pteroyl-L-glutaminsaeure Natural products C=1N=C2NC(N)=NC(=O)C2=NC=1CNC1=CC=C(C(=O)NC(CCC(O)=O)C(O)=O)C=C1 OVBPIULPVIDEAO-UHFFFAOYSA-N 0.000 description 1
- 235000010627 Phaseolus vulgaris Nutrition 0.000 description 1
- 244000046052 Phaseolus vulgaris Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000006838 adverse reaction Effects 0.000 description 1
- 235000008429 bread Nutrition 0.000 description 1
- 210000000481 breast Anatomy 0.000 description 1
- 239000000969 carrier Substances 0.000 description 1
- 235000013339 cereals Nutrition 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 235000009508 confectionery Nutrition 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 235000005289 dietary characteristics Nutrition 0.000 description 1
- 230000000378 dietary effect Effects 0.000 description 1
- 208000035475 disorder Diseases 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 229960000304 folic acid Drugs 0.000 description 1
- 235000019152 folic acid Nutrition 0.000 description 1
- 239000011724 folic acid Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 150000002632 lipids Chemical class 0.000 description 1
- 235000021056 liquid food Nutrition 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000011785 micronutrient Substances 0.000 description 1
- 235000013369 micronutrients Nutrition 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 235000021395 porridge Nutrition 0.000 description 1
- 229920001592 potato starch Polymers 0.000 description 1
- 102000004169 proteins and genes Human genes 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001225 therapeutic effect Effects 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
- G16H20/60—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to nutrition control, e.g. diets
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/022—Knowledge engineering; Knowledge acquisition
- G06N5/025—Extracting rules from data
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
- G16H10/60—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Theoretical Computer Science (AREA)
- Public Health (AREA)
- Primary Health Care (AREA)
- General Health & Medical Sciences (AREA)
- Epidemiology (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Animal Behavior & Ethology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Pathology (AREA)
- Nutrition Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The disclosure provides a personalized nutrition scheme recommendation method and system for gestational diabetics, comprising the steps of obtaining illness state portrait information of the gestational diabetics and diet preference information of the gestational diabetics, and determining the recipe interest degree of the gestational diabetics for a preset nutrition recipe based on the illness state portrait information and the diet preference information; extracting nutrition knowledge of different data sources in the nutrition field corresponding to the gestational diabetes mellitus patient, carrying out entity alignment on the nutrition knowledge, and then fusing the nutrition knowledge of the different data sources to construct a nutrition knowledge graph corresponding to the gestational diabetes mellitus; and vectorizing the nutrition knowledge maps, determining the map similarity of different nutrition knowledge maps, fusing the map similarity and the recipe interestingness, and generating recipe recommendation information. The method disclosed by the invention can accurately recommend the personalized nutrition scheme.
Description
Technical Field
The disclosure relates to the technical field of intelligent analysis, in particular to a personalized nutrition scheme recommendation method and system for gestational diabetes patients.
Background
The traditional diabetes diet recommendation algorithm faces more and more patient data and cannot match the relationship between the patient body indexes and recommended foods accurately, on one hand, the sparseness problem of user project data is solved, on the other hand, the cold start problem is solved, and as some new articles are never interacted with the user, the recommendation algorithm cannot recommend the new articles, and therefore the user cannot explore a large number of brand new articles.
Disclosure of Invention
The embodiment of the disclosure provides a personalized nutrition scheme recommendation method and a personalized nutrition scheme recommendation system for gestational diabetics, which can at least solve part of problems in the prior art, namely that the relationship between physical indexes of the patients and recommended foods cannot be accurately matched.
In a first aspect of embodiments of the present disclosure,
providing a personalized nutrition scheme recommendation method for gestational diabetes patients, comprising the following steps:
acquiring illness state portrait information of a gestational diabetes patient and diet preference information of the gestational diabetes patient, and determining the recipe interest degree of the gestational diabetes patient for a preset nutrition recipe based on the illness state portrait information and the diet preference information;
Extracting nutrition knowledge of different data sources in the nutrition field corresponding to the gestational diabetes mellitus patient, carrying out entity alignment on the nutrition knowledge, and then fusing the nutrition knowledge of the different data sources to construct a nutrition knowledge graph corresponding to the gestational diabetes mellitus;
and vectorizing the nutrition knowledge maps, determining the map similarity of different nutrition knowledge maps, fusing the map similarity and the recipe interestingness, and generating recipe recommendation information.
In an alternative embodiment of the present invention,
the determining, based on the condition portrayal information and the diet preference information, a recipe interest level of the gestational diabetes patient in a preset nutritional recipe includes:
determining an edible recipe of the gestational diabetes patient in the preset nutrition recipe based on the illness state portrait information and the preset nutrition recipe;
determining a diet preference factor contained in the edible recipe based on the diet preference information, and setting a preference weight coefficient according to the diet preference factor;
and constructing a scoring matrix by combining historical scoring information of the gestational diabetes mellitus patient, and determining the recipe interest degree of the gestational diabetes mellitus patient for a preset nutrition recipe according to the scoring matrix, the diet preference factor and the preference weight coefficient.
In an alternative embodiment of the present invention,
combining the historical scoring information of the gestational diabetes patient, constructing a scoring matrix, and determining the recipe interest degree of the gestational diabetes patient for the preset nutrition recipe according to the scoring matrix, the diet preference factor and the preference weight coefficient comprises:
wherein INS (i, j) represents the recipe interest degree of the gestational diabetes patient in a preset nutrition recipe, M, N represents the number of the gestational diabetes patients and the number of edible recipes, s respectively ij A diet preference factor representing the gestational diabetes patient i on an edible recipe j,representing the preference weight coefficient, W, corresponding to the dietary preference factor con (j) Representing a set of edible recipes, R mn Representing the scoring matrix, r mn Representing the score of the gestational diabetes patient m on the edible recipe n.
In an alternative embodiment of the present invention,
the construction of the nutrition knowledge graph corresponding to gestational diabetes comprises the following steps:
extracting entity information and relation information from nutritional knowledge of different data sources based on a pre-trained knowledge extraction model, wherein the knowledge extraction model is constructed based on a plurality of neural network models and is used for extracting the entity information and the relation information of input information of the knowledge extraction model;
Carrying out knowledge fusion through a pre-trained knowledge fusion model based on the extracted entity information and relationship information, and determining a knowledge triplet corresponding to the entity information and the relationship information, wherein the knowledge fusion model is constructed based on a classification model and is used for classifying the entity information and the relationship information;
and taking the knowledge triples as graph nodes, and taking the relations among different knowledge triples as connecting edges to construct the nutrition knowledge graph.
In an alternative embodiment of the present invention,
the knowledge extraction model before extracting entity information and relationship information from the nutritional knowledge of different data sources, the method further comprises training the knowledge extraction model:
respectively establishing a first feature template taking part of speech as a feature and a second feature template taking a context window as a feature, constructing a first training corpus according to the first feature template, and constructing a second training corpus according to the second feature template;
respectively training a knowledge extraction model to be trained by using the first training corpus and the second training corpus, determining a first training knowledge extraction model corresponding to the first training corpus and a second training knowledge extraction model corresponding to the second training corpus, respectively marking and predicting a pre-obtained unlabeled corpus by using the first training knowledge extraction model and the second training knowledge extraction model, and determining a first marking and predicting credibility corresponding to a marking and predicting result of the first training knowledge extraction model and a second marking and predicting credibility corresponding to a marking and predicting result of the second training knowledge extraction model;
The following operations are iteratively executed until a preset iteration threshold is reached:
if the reliability of the first annotation prediction is higher than a first preset annotation threshold, adding a first training corpus corresponding to the first preset annotation threshold into the second training corpus; or alternatively
If the second annotation prediction reliability is higher than a second preset annotation threshold, adding a second training corpus corresponding to the second annotation threshold higher than the second preset annotation threshold into the first training corpus.
In an alternative embodiment of the present invention,
before knowledge fusion is performed through the pre-trained knowledge fusion model, the method further comprises training the knowledge fusion model:
determining a corresponding first negative example score of each head entity, a corresponding second negative example score of each tail entity, a corresponding third negative example score of each head entity and a corresponding fourth negative example score of each tail entity in the plurality of initial positive example triples through a preset scoring function;
according to the first negative example score, the second negative example score, the third negative example score and the fourth negative example score, a first probability distribution value is distributed to the initial negative example triples and a second probability distribution value is distributed to the initial positive example triples through a preset classification function;
And based on the plurality of initial negative-example triples and the first probability distribution value, the plurality of initial positive-example triples and the second probability distribution value are used for iteratively optimizing a loss function of the knowledge fusion model to be trained through a back propagation optimization algorithm until the number of iterations is reached or a preset convergence condition is met.
In an alternative embodiment of the present invention,
the step of vectorizing the nutrition knowledge maps, the step of determining the similarity of the maps of different nutrition knowledge maps, and the step of fusing the similarity of the maps and the interest of the recipes comprises the following steps:
the similarity of the different nutrition knowledge maps is determined as shown in the following formula:
wherein Sim is 2 (J) Representing the similarity of the atlas, d and h respectively represent a direct relation coefficient and an indirect relation coefficient, r represents a weight coefficient corresponding to the indirect relation, F j 、F average A vector set respectively representing the score of the gestational diabetes patient to the nutrition knowledge graph j and the average score of the gestational diabetes patient to all the nutrition knowledge graphs;
and fusing the map similarity and the recipe interestingness as shown in the following formula:
wherein G is r And (3) representing a recommended value score, wherein A (J) represents a set of intersection of two maps in all nutrition knowledge maps, time (J) and Hot (J) respectively represent a timeliness weight and a heat weight, and INS (i, J) represents a recipe interest degree of the gestational diabetes patient in a preset nutrition recipe.
In a second aspect of the embodiments of the present disclosure,
there is provided a personalized nutritional regimen recommendation system for gestational diabetics, comprising:
a first unit for acquiring illness state portrait information of a gestational diabetes patient and diet preference information of the gestational diabetes patient, and determining a recipe interest degree of the gestational diabetes patient for a preset nutrition recipe based on the illness state portrait information and the diet preference information;
the second unit is used for extracting nutrition knowledge of different data sources in the nutrition field corresponding to the gestational diabetes mellitus patient, carrying out entity alignment on the nutrition knowledge, and then fusing the nutrition knowledge of the different data sources to construct a nutrition knowledge graph corresponding to the gestational diabetes mellitus patient;
and the third unit is used for vectorizing the nutrition knowledge maps, determining the similarity of the maps of different nutrition knowledge maps, fusing the similarity of the maps and the interest of the recipes, and generating recipe recommendation information.
In a third aspect of the embodiments of the present disclosure,
there is provided a personalized nutritional regimen recommendation device for gestational diabetics, comprising:
a processor;
a memory for storing processor-executable instructions;
Wherein the processor is configured to invoke the instructions stored in the memory to perform the method described previously.
In a fourth aspect of embodiments of the present disclosure,
there is provided a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the method as described above.
The personalized nutrition scheme recommendation method for gestational diabetes patients of the present disclosure,
in practical application, the ratio of the diet preference factors in the final diet interestingness can be further improved by setting the preference weight coefficient corresponding to the diet preference factors, so that the recommended diet not only meets the health requirements of gestational diabetes patients, but also accords with the diet habits of gestational diabetes patients, fully acquires user information, accurately determines the real-time information requirements and preferences of users, and fully meets the user requirements;
according to the method, only a small amount of marked corpus is needed for primary training in the training process, and the corpus expansion training set with high reliability is added through iterative training, so that the manpower and material resources consumed in the corpus marking link can be greatly reduced. Furthermore, because the two feature sets constructed are independent of each other, the two models trained on the independent feature sets can also be said to be independent of each other. In the training process, the two models exchange the labeled results with high reliability to each other, and in the process of exchanging labeled corpus with high reliability, the process of learning the identification features of each model is also carried out, so that the identification performance of the two models can be improved along with iteration, and finally, the parameters of the two models are fused to complete the training of the models, so that the accuracy of model identification is improved;
According to the application, by projecting the entities or the relations into the space with low dimensionality, semantic information among the entities can be effectively utilized, semantic information of the high-frequency entities can be utilized to help the low-frequency entity representation, and by modeling the relation in each dimensionality of the head and tail entities, the information of the entities and the relation can be effectively interacted; by setting the diagonal matrix function, the head and tail entities only interact in one dimension, so that the overfitting phenomenon can be effectively reduced; and the complex representation function is combined, the representation of the entity and the relationship is expanded to the complex field, and the asymmetric relationship can be effectively represented.
Drawings
FIG. 1 is a flow chart of a personalized nutritional regimen recommendation method for gestational diabetics in accordance with an embodiment of the present disclosure;
FIG. 2 is a schematic diagram of a food subcategory according to an embodiment of the present disclosure;
FIG. 3 is a schematic illustration of a nutrient class of an embodiment of the present disclosure;
fig. 4 is a schematic diagram of a personalized nutrition program recommendation system for gestational diabetics according to an embodiment of the disclosure.
Detailed Description
For the purposes of making the objects, technical solutions and advantages of the embodiments of the present disclosure more apparent, the technical solutions of the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present disclosure, and it is apparent that the described embodiments are only some embodiments of the present disclosure, not all embodiments. Based on the embodiments in this disclosure, all other embodiments that a person of ordinary skill in the art would obtain without making any inventive effort are within the scope of protection of this disclosure.
The technical scheme of the present disclosure is described in detail below with specific examples. The following embodiments may be combined with each other, and some embodiments may not be repeated for the same or similar concepts or processes.
Fig. 1 is a flow chart of a personalized nutrition program recommendation method for gestational diabetes patients according to an embodiment of the present disclosure, as shown in fig. 1, the method includes:
s101, acquiring illness state portrait information of a gestational diabetes patient and diet preference information of the gestational diabetes patient, and determining the recipe interest degree of the gestational diabetes patient for a preset nutrition recipe based on the illness state portrait information and the diet preference information;
in practical application, because the acquired user information is not comprehensive and abundant enough, the analysis technology is imperfect and other factors influence, the real-time information requirement and preference of the user are difficult to accurately determine, and therefore, the user requirement is difficult to comprehensively meet.
For example, the condition image information may include a type of diabetes patient, a blood sugar value, a blood fat value, etc., and may further include a diet feature point, a exercise feature point, etc. And extracting various pathological characteristic points (such as type characteristic points, blood sugar value characteristic points and blood fat value characteristic points of diabetics), various dietary characteristic points (porridge characteristic points, steamed stuffed bun characteristic points and steamed bread characteristic points) and various movement characteristic points (running characteristic points and slow walking characteristic points) of the diabetics from the multi-source heterogeneous data.
The condition portrait is the characteristic of describing the condition of the user and is the premise of realizing information recommendation. The application constructs the condition portrait information of the user through the basic characteristics of the user, the diet information facing the user requirement and the user interaction information.
The user basic characteristics can comprise user ID, gender, age, illness history and the like, and the diet information facing the user demand can comprise diet information which can be eaten aiming at illness state or physical state of the user; the user interaction information can comprise behavior information such as information release, reading, collection, interaction participation and the like of the user in the social software platform. It should be noted that, the method for constructing the patient image information of the user may refer to the existing method, which is not limited in the embodiment of the present application.
For example, the dietary preference information of the gestational diabetes patient may include tag information of the user acquired from a third party platform, such as a take-out platform, optionally, if the take-out types selected by the user in the take-out platform are all of a spicy and biased type, the corresponding tags of the user may be acquired from the take-out platform as the dietary preference information of the gestational diabetes patient.
In an alternative embodiment of the present invention,
the determining, based on the condition portrayal information and the diet preference information, a recipe interest level of the gestational diabetes patient in a preset nutritional recipe includes:
determining an edible recipe of the gestational diabetes patient in the preset nutrition recipe based on the illness state portrait information and the preset nutrition recipe;
determining a diet preference factor contained in the edible recipe based on the diet preference information, and setting a preference weight coefficient according to the diet preference factor;
and constructing a scoring matrix by combining historical scoring information of the gestational diabetes mellitus patient, and determining the recipe interest degree of the gestational diabetes mellitus patient for a preset nutrition recipe according to the scoring matrix, the diet preference factor and the preference weight coefficient.
Illustratively, the preset nutritional recipes comprise a plurality of types of recipes which can be used for different types of users, the edible recipes of the gestational diabetes patients are screened from the preset nutritional recipes by combining with disease images, for example, in a complete breast feeding stage, the daily intake of calories is increased by 400-500kCal, the daily intake of a minimum of 175 g of carbohydrates, the daily addition of 400ug of folic acid, the addition of Fe and Ca preparations or a micro-nutrient compound preparation suitable for the gestation period are carried out, and the recipes of interest for the gestational diabetes patients are further screened on the basis of the edible recipes;
the diet preference information can comprise various diet preference factors, in particular, the diet preference factors can comprise spicy, sweet, cooked wheaten food, liquid food, fried food and other taste information, and if the diet comprises the diet preference factors, the user is biased to be interested in the diet;
in addition, the historical scoring information of the gestational diabetes patient can be combined to construct a scoring matrix, and in particular, the scoring matrix can be shown in the following formula:
wherein R is mn Representing the scoring matrix, r mn Representing the score of the gestational diabetes patient m on the edible recipe n.
In an alternative embodiment of the present invention,
Combining the historical scoring information of the gestational diabetes patient, constructing a scoring matrix, and determining the recipe interest degree of the gestational diabetes patient for the preset nutrition recipe according to the scoring matrix, the diet preference factor and the preference weight coefficient comprises:
wherein INS (i, j) represents the recipe interest degree of the gestational diabetes patient in a preset nutrition recipe, M, N represents the number of the gestational diabetes patients and the number of edible recipes, s respectively ij A diet preference factor representing the gestational diabetes patient i on an edible recipe j,representing the preference weight coefficient, W, corresponding to the dietary preference factor con (j) Representing a set of edible recipes, R mn Representing the scoring matrix, r mn Representing the score of the gestational diabetes patient m on the edible recipe n.
The ratio of the diet preference factors in the final diet interestingness can be further improved by setting the preference weight coefficient corresponding to the diet preference factors, so that the recommended diet not only meets the health requirements of gestational diabetes patients, but also accords with the diet habits of the gestational diabetes patients, and the user experience is improved.
S102, extracting nutrition knowledge of different data sources in the nutrition field corresponding to the gestational diabetes mellitus patient, and fusing the nutrition knowledge of the different data sources after carrying out entity alignment on the nutrition knowledge to construct a nutrition knowledge graph corresponding to the gestational diabetes mellitus;
by way of example, the knowledge graph can well describe entities and relations in the objective world, and the problem that a machine learning algorithm and a deep learning algorithm are weak in describing ability is solved. In the knowledge graph, objective things in the real world are called entities, such as objects, figures, institutions, and the like, and associations between two entities are called relationships, such as relationships between students and schools. The entities and relationships are stored in the knowledge graph in a manner of triplet < head entity, relationship, tail entity >. The knowledge graph converts information in mass data into a structured triplet, so that the knowledge graph can effectively store and use the data while describing the data.
With the improvement of living standard and the continuous progress and development of society, people find that diet plays a very important role in physical health. Pre-diabetic patients can reduce the incidence of diabetes for a long period of time by intervening in lifestyles such as diet and exercise. The most important in the diet is the balance of nutrient elements, and the content of suitable or unsuitable nutrient elements of people with different physical conditions is different. The nutritionist can provide different dietary or sports advice to different patients according to his own learned nutriology related knowledge and experience during diagnosis, and although the nutritionist can give related advice of daily intake of nutritional ingredients, the physical condition of each person is constantly changing for most people, and especially for gestational diabetics, it is troublesome to frequently seek advice of the nutritionist. Therefore, it is necessary to label the knowledge in the nutrition professional book and extract a great amount of nutrition, diet and group disease related data in the internet to construct a knowledge graph in the nutrition field, so as to provide timely and accurate diet advice for users.
In an alternative embodiment of the present invention,
the construction of the nutrition knowledge graph corresponding to gestational diabetes comprises the following steps:
extracting entity information and relation information from nutritional knowledge of different data sources based on a pre-trained knowledge extraction model, wherein the knowledge extraction model is constructed based on a plurality of neural network models and is used for extracting the entity information and the relation information of input information of the knowledge extraction model;
carrying out knowledge fusion through a pre-trained knowledge fusion model based on the extracted entity information and relationship information, and determining a knowledge triplet corresponding to the entity information and the relationship information, wherein the knowledge fusion model is constructed based on a classification model and is used for classifying the entity information and the relationship information;
and taking the knowledge triples as graph nodes, and taking the relations among different knowledge triples as connecting edges to construct the nutrition knowledge graph.
The most central links in the construction process of the knowledge graph are knowledge extraction and knowledge fusion, wherein the knowledge fusion is divided into entity information fusion and relationship information fusion. Structured data is normalized data, and knowledge in the structured data can be easily identified only according to the characteristics of the structured data during knowledge extraction. The semi-structured and unstructured data have poorer normalization than structured data due to different carriers and non-uniform encoding formats, so that knowledge extraction is specially performed according to the respective characteristics.
Entity extraction is also called named entity recognition, and the accuracy of entity extraction directly affects the quality and efficiency of knowledge extraction, so entity extraction is the basis and key of knowledge graph construction and knowledge extraction.
Further, by extracting entity information and relationship information in structured, semi-structured, and unstructured corpus text, the problems of the extracted knowledge are as follows: on the one hand, the semi-structured or unstructured corpus text has information redundancy; on the other hand, the data are mutually independent, and the semantic relation of the upper and lower layers is not obvious, so that the extracted knowledge in the semi-structured or unstructured corpus text cannot be directly used, and the extracted knowledge needs to be tidied through knowledge fusion.
In an alternative embodiment of the present invention,
extracting entity information and relation information from nutritional knowledge of different data sources based on a pre-trained knowledge extraction model, wherein the knowledge extraction model is constructed based on a plurality of neural network models and is used for extracting the entity information and the relation information of input information of the knowledge extraction model;
before the knowledge extraction model extracts entity information and relationship information from the nutritional knowledge of the different data sources, the method further comprises training the knowledge extraction model:
Respectively establishing a first feature template taking part of speech as a feature and a second feature template taking a context window as a feature, constructing a first training corpus according to the first feature template, and constructing a second training corpus according to the second feature template;
respectively training a knowledge extraction model to be trained by using the first training corpus and the second training corpus, determining a first training knowledge extraction model corresponding to the first training corpus and a second training knowledge extraction model corresponding to the second training corpus, respectively marking and predicting a pre-obtained unlabeled corpus by using the first training knowledge extraction model and the second training knowledge extraction model, and determining a first marking and predicting credibility corresponding to a marking and predicting result of the first training knowledge extraction model and a second marking and predicting credibility corresponding to a marking and predicting result of the second training knowledge extraction model;
the following operations are iteratively executed until a preset iteration threshold is reached:
if the reliability of the first annotation prediction is higher than a first preset annotation threshold, adding a first training corpus corresponding to the first preset annotation threshold into the second training corpus; or alternatively
If the second annotation prediction reliability is higher than a second preset annotation threshold, adding a second training corpus corresponding to the second annotation threshold higher than the second preset annotation threshold into the first training corpus.
By way of example, a feature set can be formulated according to a sequence labeling task scene, two independent features are selected from the feature set, and a corresponding feature template is established.
Inputting a large number of unlabeled corpora into a model obtained by primary training, carrying out labeling prediction, and evaluating the reliability of a prediction result; and finally, adding the labeling corpus with high reliability in the prediction result into the training corpus of the other party, and continuing training the model by using the new corpus after expansion. This process repeats the generation until training is stopped after the contract conditions are triggered.
According to the method, only a small amount of marked corpus is needed for primary training in the training process, and the corpus expansion training set with high reliability is added through iterative training, so that the manpower and material resources consumed in the corpus marking link can be greatly reduced. Furthermore, because the two feature sets constructed are independent of each other, the two models trained on the independent feature sets can also be said to be independent of each other. In the training process, the two models exchange the labeled results with high reliability to each other, and in the process of exchanging labeled corpus with high reliability, the process of learning the identification features of each model is also carried out, so that the identification performance of the two models can be improved along with iteration, and finally, the parameters of the two models are fused to complete the training of the models, and the accuracy of model identification is improved.
In an alternative embodiment of the present invention,
before knowledge fusion is performed through the pre-trained knowledge fusion model, the method further comprises training the knowledge fusion model:
performing knowledge scoring on the extracted entity information and the relationship information, constructing the entity information and the relationship information, of which the knowledge scoring exceeds a preset score threshold, into a knowledge fusion training set, wherein performing knowledge scoring on the extracted entity information and the relationship information comprises:
wherein S is h,r,t Knowledge scores representing entity information and relationship information, N representing the number of all training triples, N r Representing the number, N, of relation information r in the training triplet h,r Representing the number of co-occurrences of the head entity h and the relation r in the training triplet, N h Representing the number of head entities h in the training triplet, N r,t Representing the number of co-occurrences of the relation r and the tail entity t in the training triplet;
and classifying the entity information and the relation information of the knowledge fusion training set through a classification function in a pre-trained knowledge fusion model, determining the spatial distance between the entity information and the relation information, and constructing a knowledge triplet by taking the information with the spatial distance smaller than a preset distance threshold as the same category.
By way of example, the respective entity definition of the extracted nutritional knowledge of the different data sources may be based on the type and relationship of the entity present in the knowledge base in combination with the analysis of the data in the professional book and website, under the direction of the nutritional specialist, divide the nutritional entity into eight major categories, including food, dish, nutrient, disease, symptom, organ, function and crowd, around three aspects of nutrition, medical and crowd. Fig. 2 is a schematic diagram of the food subcategories according to the embodiment of the present disclosure, and referring to the classification of food in the "chinese food composition table", the food subcategories may be divided into 16 subclasses of cereal, potato starch, dried bean, oil and fat, etc., while fig. 3 is a schematic diagram of the nutrient class according to the embodiment of the present disclosure, and the nutrient class may be further divided into 7 subclasses of carbohydrate, lipid, protein, etc.
Defining the relationship of the extracted nutritional knowledge of the different data sources, respectively, may include: at least one of contraindication, suitable eating and side effects;
wherein, the tabu eating relationship comprises: "treatment-food", "crowd-food", "disease-food", "symptomatic food" and the like mean that a person cannot or is unsuitable to eat certain foods by himself because of a certain special state of the person, if eating is uncertain, an abnormal reaction occurs, but a potential adverse reaction risk exists, and safety guarantee is provided for future dietary recommendation generation recipes.
Wherein, suitable eating relationships include: "disease-food", "symptom-food", "crowd-food" means that when a human body is in a specific state, eating a certain food is beneficial to improving the current state or beneficial to health. The method has the significance that when the diet recommendation system generates the recipes, more influencing factors can be considered, and more personalized recipe recommendation can be obtained.
Wherein, the side effect relationship includes: "treatment-disorder" and "treatment-symptom" refer to a negative disease or symptom that a patient may experience in addition to the intended therapeutic effect when receiving a certain treatment or taking a certain medication, by which means a possibility of developing more food and disease, symptom and treatment relationships for future dietary recommendations is provided.
Entity-to-entity relationships are extracted in structured, semi-structured, and unstructured corpus text by related techniques and methods, but the extracted knowledge has the following problems: in one aspect, semi-structured or unstructured corpus text is subject to redundancy of information: on the other hand, the data are mutually independent, and the semantic relation of the upper and lower layers is not obvious, so that the extracted knowledge in the semi-structured or unstructured corpus text cannot be directly used, and the extracted knowledge needs to be tidied through knowledge fusion.
In an alternative embodiment of the present invention,
before knowledge fusion is performed through the pre-trained knowledge fusion model, the method further comprises training the knowledge fusion model:
determining a corresponding first negative example score of each head entity, a corresponding second negative example score of each tail entity, a corresponding third negative example score of each head entity and a corresponding fourth negative example score of each tail entity in the plurality of initial positive example triples through a preset scoring function;
according to the first negative example score, the second negative example score, the third negative example score and the fourth negative example score, a first probability distribution value is distributed to the initial negative example triples and a second probability distribution value is distributed to the initial positive example triples through a preset classification function;
and based on the plurality of initial negative-example triples and the first probability distribution value, the plurality of initial positive-example triples and the second probability distribution value are used for iteratively optimizing a loss function of the knowledge fusion model to be trained through a back propagation optimization algorithm until the number of iterations is reached or a preset convergence condition is met.
Before constructing the knowledge graph, the method further comprises: carrying out knowledge graph link prediction for filling incomplete triples;
the aim of knowledge graph link prediction is to fill incomplete triplets, the link prediction task mainly examines the triplets as a whole, the relation between the real object and the relation,
determining semantic representations of the head entity, the tail entity and the entity relationship according to the vector representations of the head entity, the vector representations of the tail entity, a preset complex representation function, a diagonal matrix function and the matrix representations of the head entity, the tail entity and the entity relationship;
wherein f r (h, t) represents semantic representation of head entity h, tail entity t and entity relationship r, D represents vector quantity of head entity and tail entity, re { } represents complex representation function, diag represents diagonal matrix function,matrix representation corresponding to the entity relation r representing the i-th head entity and the j-th tail entity,/->Vector representation representing the i-th header entity, a +.>A vector representation representing a j-th tail entity;
by projecting entities or relationships into a low-dimensional space, we can efficiently use semantic information between entities and can use semantic information of high frequency entities to help low frequency entity representations. By modeling each dimension of the relationship in the head-to-tail entity, each relationship r has parameters of O (d 2) dimensions, and the information of the entity and the relationship can be effectively interacted; meanwhile, in order to avoid the phenomenon that the model is more complex and easy to be over-fitted along with the increase of the dimension of the relation matrix, the head and tail entities only interact in one dimension by setting the diagonal matrix function, so that the over-fitting phenomenon can be effectively reduced; and the complex representation function is combined, the representation of the entity and the relationship is expanded to the complex field, and the asymmetric relationship can be effectively represented.
And S103, vectorizing the nutrition knowledge maps, determining the similarity of the maps of different nutrition knowledge maps, and fusing the similarity of the maps and the interest degree of the recipes to generate recipe recommendation information.
In an alternative embodiment of the present invention,
the step of vectorizing the nutrition knowledge maps, the step of determining the similarity of the maps of different nutrition knowledge maps, and the step of fusing the similarity of the maps and the interest of the recipes comprises the following steps:
the similarity of the different nutrition knowledge maps is determined as shown in the following formula:
wherein Sim is 2 (J) Representing the similarity of the atlas, d and h respectively represent a direct relation coefficient and an indirect relation coefficient, r represents a weight coefficient corresponding to the indirect relation, F j 、F average A vector set respectively representing the score of the gestational diabetes patient to the nutrition knowledge graph j and the average score of the gestational diabetes patient to all the nutrition knowledge graphs;
and fusing the map similarity and the recipe interestingness as shown in the following formula:
wherein G is r And (3) representing a recommended value score, wherein A (J) represents a set of intersection of two maps in all nutrition knowledge maps, time (J) and Hot (J) respectively represent a timeliness weight and a heat weight, and INS (i, J) represents a recipe interest degree of the gestational diabetes patient in a preset nutrition recipe.
Alternatively, the text may be ranked based on the calculated recommendation value scores, and finally, recipes corresponding to the first k recommendation value scores are selected to form a recommendation list.
For example, for any piece of information to be recommended, whether the information is recommended or not is influenced by the similarity between the information and the content of the current information or is related to whether the information has the characteristic of attracting users to read, so that the two factors are comprehensively considered to calculate the total weight of the recommendation degree. The timeliness weight and the heat weight can be used for indicating timeliness and heat of information in the content recommendation process, and accuracy of information recommendation can be further improved. In addition, the direct relationship and the indirect relationship can be used for indicating the distance degree between the information, wherein the direct relationship indicates that the association between the two information is relatively close, the indirect relationship indicates that the association between the two information is relatively far, and the association between the two information can be further enhanced by distributing the corresponding weight value for the indirect relationship.
According to the method, the semantic similarity among the recipes is calculated by vectorizing the recipes in the knowledge graph, and the semantic similarity is fused with the similarity based on the user behaviors, so that the final recipe similarity is calculated, and further a recommendation list is generated, the effect of combining external scoring and internal information is achieved, the defect that the internal information of the recipes is ignored by the existing recommendation algorithm can be overcome, the problems of data sparsity and cold start in the recommendation process are solved, and the accuracy and reasonability of recommendation are improved.
In a second aspect of the embodiments of the present disclosure,
providing a personalized nutrition scheme recommendation system for gestational diabetics, fig. 4 is a schematic structural diagram of a personalized nutrition scheme recommendation system for gestational diabetics according to an embodiment of the disclosure, including:
a first unit for acquiring illness state portrait information of a gestational diabetes patient and diet preference information of the gestational diabetes patient, and determining a recipe interest degree of the gestational diabetes patient for a preset nutrition recipe based on the illness state portrait information and the diet preference information;
the second unit is used for extracting nutrition knowledge of different data sources in the nutrition field corresponding to the gestational diabetes mellitus patient, carrying out entity alignment on the nutrition knowledge, and then fusing the nutrition knowledge of the different data sources to construct a nutrition knowledge graph corresponding to the gestational diabetes mellitus patient;
and the third unit is used for vectorizing the nutrition knowledge maps, determining the similarity of the maps of different nutrition knowledge maps, fusing the similarity of the maps and the interest of the recipes, and generating recipe recommendation information.
In a third aspect of the embodiments of the present disclosure,
there is provided a personalized nutritional regimen recommendation device for gestational diabetics, comprising:
A processor;
a memory for storing processor-executable instructions;
wherein the processor is configured to invoke the instructions stored in the memory to perform the method described previously.
In a fourth aspect of embodiments of the present disclosure,
there is provided a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the method as described above.
The present invention may be a method, apparatus, system, and/or computer program product. The computer program product may include a computer readable storage medium having computer readable program instructions embodied thereon for performing various aspects of the present invention.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present disclosure, and not for limiting the same; although the present disclosure has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present disclosure.
Claims (5)
1. A personalized nutritional regimen recommendation method for gestational diabetics, comprising:
acquiring illness state portrait information of a gestational diabetes patient and diet preference information of the gestational diabetes patient, and determining the recipe interest degree of the gestational diabetes patient for a preset nutrition recipe based on the illness state portrait information and the diet preference information;
the determining, based on the condition portrayal information and the diet preference information, a recipe interest level of the gestational diabetes patient in a preset nutritional recipe includes:
determining an edible recipe of the gestational diabetes patient in the preset nutrition recipe based on the illness state portrait information and the preset nutrition recipe;
determining a diet preference factor contained in the edible recipe based on the diet preference information, and setting a preference weight coefficient according to the diet preference factor;
combining historical scoring information of the gestational diabetes mellitus patient, constructing a scoring matrix, and determining the recipe interest degree of the gestational diabetes mellitus patient for a preset nutrition recipe according to the scoring matrix, the diet preference factors and the preference weight coefficients;
Combining the historical scoring information of the gestational diabetes patient, constructing a scoring matrix, and determining the recipe interest degree of the gestational diabetes patient for the preset nutrition recipe according to the scoring matrix, the diet preference factor and the preference weight coefficient comprises:
;
;
wherein,INS(i,j)representing the recipe interest of the gestational diabetes patient in the edible recipe,M、Nthe number of diabetics in gestation and the number of edible recipes are respectively represented,s ij representing the gestational diabetes patientiFor edible recipesjIs a function of the dietary preference factor of (a),representing preference weighting coefficients corresponding to the dietary preference factors,W con (j)a set of edible recipes is represented,R ij the scoring matrix is represented by a matrix of scores,r ij representing the gestational diabetesPatient's healthiFor edible recipesjIs a score of (2);
extracting nutrition knowledge of different data sources in the nutrition field corresponding to the gestational diabetes mellitus patient, carrying out entity alignment on the nutrition knowledge, and then fusing the nutrition knowledge of the different data sources to construct a nutrition knowledge graph corresponding to the gestational diabetes mellitus;
vectorizing the nutrition knowledge maps, determining the map similarity of different nutrition knowledge maps, fusing the map similarity and the recipe interestingness, and generating recipe recommendation information;
The step of vectorizing the nutrition knowledge maps, the step of determining the similarity of the maps of different nutrition knowledge maps, and the step of fusing the similarity of the maps and the interest of the recipes comprises the following steps:
the similarity of the different nutrition knowledge maps is determined as shown in the following formula:
;
wherein,Sim 2 (J)the similarity of the map is represented by a graph,d()、h()respectively represent a direct off coefficient and an indirect off coefficient,rthe weight coefficient corresponding to the indirect relation is represented,F k 、F average respectively represent the nutrition knowledge graph of gestational diabetes patientskA vector set of scores of a gestational diabetes patient on an average score of all nutritional knowledge maps;
and fusing the map similarity and the recipe interestingness as shown in the following formula:
;
wherein,G r a recommendation value score is represented and,A(J)represents the set of intersection of two maps in all nutrition knowledge maps,Time (j)、Hot(j)respectively representing timeliness weight and heat weight,INS(i,j)representing a recipe interest level of the gestational diabetes patient in an edible recipe;
the construction of the nutrition knowledge graph corresponding to gestational diabetes comprises the following steps:
extracting entity information and relation information from nutritional knowledge of different data sources based on a pre-trained knowledge extraction model, wherein the knowledge extraction model is constructed based on a plurality of neural network models and is used for extracting the entity information and the relation information of input information of the knowledge extraction model;
Carrying out knowledge fusion through a pre-trained knowledge fusion model based on the extracted entity information and relationship information, and determining a knowledge triplet corresponding to the entity information and the relationship information, wherein the knowledge fusion model is constructed based on a classification model and is used for classifying the entity information and the relationship information;
the knowledge triples are used as graph nodes, the relations among different knowledge triples are used as connecting edges, and the nutrition knowledge graph is constructed;
before the knowledge extraction model extracts entity information and relationship information from the nutritional knowledge of the different data sources, the method further comprises training the knowledge extraction model:
respectively establishing a first feature template taking part of speech as a feature and a second feature template taking a context window as a feature, constructing a first training corpus according to the first feature template, and constructing a second training corpus according to the second feature template;
respectively training a knowledge extraction model to be trained by using the first training corpus and the second training corpus, determining a first training knowledge extraction model corresponding to the first training corpus and a second training knowledge extraction model corresponding to the second training corpus, respectively marking and predicting a pre-obtained unlabeled corpus by using the first training knowledge extraction model and the second training knowledge extraction model, and determining a first marking and predicting credibility corresponding to a marking and predicting result of the first training knowledge extraction model and a second marking and predicting credibility corresponding to a marking and predicting result of the second training knowledge extraction model;
The following operations are iteratively executed until a preset iteration threshold is reached:
if the reliability of the first annotation prediction is higher than a first preset annotation threshold, adding a first training corpus corresponding to the first preset annotation threshold into the second training corpus; or alternatively
If the second annotation prediction reliability is higher than a second preset annotation threshold, adding a second training corpus corresponding to the second annotation threshold higher than the second preset annotation threshold into the first training corpus.
2. The method of claim 1, wherein prior to knowledge fusion by a pre-trained knowledge fusion model, the method further comprises training the knowledge fusion model:
determining a corresponding first negative example score of each head entity, a corresponding second negative example score of each tail entity, a corresponding third negative example score of each head entity and a corresponding fourth negative example score of each tail entity in the plurality of initial positive example triples through a preset scoring function;
according to the first negative example score, the second negative example score, the third negative example score and the fourth negative example score, a first probability distribution value is distributed to the initial negative example triples and a second probability distribution value is distributed to the initial positive example triples through a preset classification function;
And based on the plurality of initial negative-example triples and the first probability distribution value, the plurality of initial positive-example triples and the second probability distribution value are used for iteratively optimizing a loss function of the knowledge fusion model to be trained through a back propagation optimization algorithm until the number of iterations is reached or a preset convergence condition is met.
3. A personalized nutritional regimen recommendation system for gestational diabetics, comprising:
a first unit for acquiring illness state portrait information of a gestational diabetes patient and diet preference information of the gestational diabetes patient, and determining a recipe interest degree of the gestational diabetes patient for a preset nutrition recipe based on the illness state portrait information and the diet preference information;
the determining, based on the condition portrayal information and the diet preference information, a recipe interest level of the gestational diabetes patient in a preset nutritional recipe includes:
determining an edible recipe of the gestational diabetes patient in the preset nutrition recipe based on the illness state portrait information and the preset nutrition recipe;
determining a diet preference factor contained in the edible recipe based on the diet preference information, and setting a preference weight coefficient according to the diet preference factor;
Combining historical scoring information of the gestational diabetes mellitus patient, constructing a scoring matrix, and determining the recipe interest degree of the gestational diabetes mellitus patient for a preset nutrition recipe according to the scoring matrix, the diet preference factors and the preference weight coefficients;
combining the historical scoring information of the gestational diabetes patient, constructing a scoring matrix, and determining the recipe interest degree of the gestational diabetes patient for the preset nutrition recipe according to the scoring matrix, the diet preference factor and the preference weight coefficient comprises:
;
;
wherein,INS(i,j)representing the recipe interest of the gestational diabetes patient in the edible recipe,M、Nthe number of diabetics in gestation and the number of edible recipes are respectively represented,s ij representing the gestational diabetes patientiFor edible recipesjIs a function of the dietary preference factor of (a),representing preference weighting coefficients corresponding to the dietary preference factors,W con (j)a set of edible recipes is represented,R ij the scoring matrix is represented by a matrix of scores,r ij representing the gestational diabetes patientiFor edible recipesjIs a score of (2);
the second unit is used for extracting nutrition knowledge of different data sources in the nutrition field corresponding to the gestational diabetes mellitus patient, carrying out entity alignment on the nutrition knowledge, and then fusing the nutrition knowledge of the different data sources to construct a nutrition knowledge graph corresponding to the gestational diabetes mellitus patient;
The third unit is used for vectorizing the nutrition knowledge maps, determining the similarity of the maps of different nutrition knowledge maps, fusing the similarity of the maps and the interest of the recipes, and generating recipe recommendation information;
the step of vectorizing the nutrition knowledge maps, the step of determining the similarity of the maps of different nutrition knowledge maps, and the step of fusing the similarity of the maps and the interest of the recipes comprises the following steps:
the similarity of the different nutrition knowledge maps is determined as shown in the following formula:
;
wherein,Sim 2 (J)the similarity of the map is represented by a graph,d()、h()respectively represent a direct off coefficient and an indirect off coefficient,rthe weight coefficient corresponding to the indirect relation is represented,F k 、F average respectively represent the nutrition knowledge graph of gestational diabetes patientskA vector set of scores of a gestational diabetes patient on an average score of all nutritional knowledge maps;
and fusing the map similarity and the recipe interestingness as shown in the following formula:
;
wherein,G r a recommendation value score is represented and,A(J)represents the set of intersection of two maps in all nutrition knowledge maps,Time (j)、Hot(j)respectively representing timeliness weight and heat weight,INS(i,j)representing a recipe interest level of the gestational diabetes patient in an edible recipe;
The second unit is further configured to: extracting entity information and relation information from nutritional knowledge of different data sources based on a pre-trained knowledge extraction model, wherein the knowledge extraction model is constructed based on a plurality of neural network models and is used for extracting the entity information and the relation information of input information of the knowledge extraction model;
carrying out knowledge fusion through a pre-trained knowledge fusion model based on the extracted entity information and relationship information, and determining a knowledge triplet corresponding to the entity information and the relationship information, wherein the knowledge fusion model is constructed based on a classification model and is used for classifying the entity information and the relationship information;
the knowledge triples are used as graph nodes, the relations among different knowledge triples are used as connecting edges, and the nutrition knowledge graph is constructed;
the second unit is further configured to train the knowledge extraction model:
respectively establishing a first feature template taking part of speech as a feature and a second feature template taking a context window as a feature, constructing a first training corpus according to the first feature template, and constructing a second training corpus according to the second feature template;
Respectively training a knowledge extraction model to be trained by using the first training corpus and the second training corpus, determining a first training knowledge extraction model corresponding to the first training corpus and a second training knowledge extraction model corresponding to the second training corpus, respectively marking and predicting a pre-obtained unlabeled corpus by using the first training knowledge extraction model and the second training knowledge extraction model, and determining a first marking and predicting credibility corresponding to a marking and predicting result of the first training knowledge extraction model and a second marking and predicting credibility corresponding to a marking and predicting result of the second training knowledge extraction model;
the following operations are iteratively executed until a preset iteration threshold is reached:
if the reliability of the first annotation prediction is higher than a first preset annotation threshold, adding a first training corpus corresponding to the first preset annotation threshold into the second training corpus; or alternatively
If the second annotation prediction reliability is higher than a second preset annotation threshold, adding a second training corpus corresponding to the second annotation threshold higher than the second preset annotation threshold into the first training corpus.
4. A personalized nutritional regimen recommendation device for gestational diabetics, comprising:
A processor;
a memory for storing processor-executable instructions;
wherein the processor is configured to invoke the instructions stored in the memory to perform the method of any of claims 1 to 2.
5. A computer readable storage medium having stored thereon computer program instructions, which when executed by a processor, implement the method of any of claims 1 to 2.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310665199.9A CN116417115B (en) | 2023-06-07 | 2023-06-07 | Personalized nutrition scheme recommendation method and system for gestational diabetes patients |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310665199.9A CN116417115B (en) | 2023-06-07 | 2023-06-07 | Personalized nutrition scheme recommendation method and system for gestational diabetes patients |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116417115A CN116417115A (en) | 2023-07-11 |
CN116417115B true CN116417115B (en) | 2023-12-01 |
Family
ID=87059651
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310665199.9A Active CN116417115B (en) | 2023-06-07 | 2023-06-07 | Personalized nutrition scheme recommendation method and system for gestational diabetes patients |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116417115B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116580817B (en) * | 2023-07-12 | 2023-09-22 | 北京四海汇智科技有限公司 | Personalized nutrition guidance method and system based on machine learning |
CN117133407B (en) * | 2023-10-26 | 2024-02-13 | 北京四海汇智科技有限公司 | Nutritional balance assessment method and system for multi-label neural network for children |
CN117558410B (en) * | 2024-01-12 | 2024-03-22 | 吉林大学 | Artificial intelligence-based diet management system and method for diabetic nephropathy patients |
Citations (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110287306A (en) * | 2019-06-26 | 2019-09-27 | 珠海格力电器股份有限公司 | Recipe recommendation method and equipment |
CN110310721A (en) * | 2018-03-27 | 2019-10-08 | 九阳股份有限公司 | A kind of method and device of knowledge based map construction recipe recommendation strategy |
CN110379486A (en) * | 2019-07-09 | 2019-10-25 | 北京大学第一医院 | A kind of gestational period blood glucose health analysis system |
CN110504019A (en) * | 2019-08-30 | 2019-11-26 | 北京妙医佳健康科技集团有限公司 | User individual dietary recommendations continued method, apparatus, electronic equipment and storage medium |
CN110609903A (en) * | 2019-08-01 | 2019-12-24 | 华为技术有限公司 | Information presentation method and device |
CN111883230A (en) * | 2019-12-18 | 2020-11-03 | 深圳数字生命研究院 | Method and device for generating diet data, storage medium and electronic device |
CN111950269A (en) * | 2020-08-21 | 2020-11-17 | 清华大学 | Text statement processing method and device, computer equipment and storage medium |
CN112528008A (en) * | 2020-12-07 | 2021-03-19 | 北京健康有益科技有限公司 | Diabetic catering method and device based on knowledge graph |
CN112614564A (en) * | 2020-12-24 | 2021-04-06 | 武汉理工大学 | Intelligent nutritional diet recommendation method based on multi-source heterogeneous data fusion |
CN112818222A (en) * | 2021-01-26 | 2021-05-18 | 吾征智能技术(北京)有限公司 | Knowledge graph-based personalized diet recommendation method and system |
CN113611393A (en) * | 2021-08-20 | 2021-11-05 | 平安国际智慧城市科技股份有限公司 | Food recommendation, device, equipment and storage medium based on artificial intelligence |
CN114429809A (en) * | 2022-01-25 | 2022-05-03 | 武汉科瓴智能科技有限公司 | Individual nutrition scheme generation method and system for diabetic patients |
CN114758749A (en) * | 2022-03-23 | 2022-07-15 | 清华大学 | Nutritional diet management map creation method and device based on gestation period |
CN114861917A (en) * | 2022-04-29 | 2022-08-05 | 华中科技大学 | Knowledge graph inference model, system and inference method for Bayesian small sample learning |
CN115280422A (en) * | 2020-12-25 | 2022-11-01 | 京东方科技集团股份有限公司 | Food recommendation method and device, storage medium and electronic equipment |
CN115602289A (en) * | 2022-10-13 | 2023-01-13 | 中南大学(Cn) | Dynamic diet recommendation method based on crowd attribute-diet knowledge graph |
CN115714001A (en) * | 2022-11-14 | 2023-02-24 | 北京航空航天大学 | Method for constructing multi-modal knowledge graph service platform for healthy diet |
CN116127089A (en) * | 2022-12-09 | 2023-05-16 | 海信视像科技股份有限公司 | Intelligent device, menu recommendation method and medium |
CN116166812A (en) * | 2022-12-12 | 2023-05-26 | 中国科学院自动化研究所 | Knowledge graph completion method and device, electronic equipment and storage medium |
-
2023
- 2023-06-07 CN CN202310665199.9A patent/CN116417115B/en active Active
Patent Citations (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110310721A (en) * | 2018-03-27 | 2019-10-08 | 九阳股份有限公司 | A kind of method and device of knowledge based map construction recipe recommendation strategy |
CN110287306A (en) * | 2019-06-26 | 2019-09-27 | 珠海格力电器股份有限公司 | Recipe recommendation method and equipment |
CN110379486A (en) * | 2019-07-09 | 2019-10-25 | 北京大学第一医院 | A kind of gestational period blood glucose health analysis system |
CN110609903A (en) * | 2019-08-01 | 2019-12-24 | 华为技术有限公司 | Information presentation method and device |
CN110504019A (en) * | 2019-08-30 | 2019-11-26 | 北京妙医佳健康科技集团有限公司 | User individual dietary recommendations continued method, apparatus, electronic equipment and storage medium |
CN111883230A (en) * | 2019-12-18 | 2020-11-03 | 深圳数字生命研究院 | Method and device for generating diet data, storage medium and electronic device |
CN111950269A (en) * | 2020-08-21 | 2020-11-17 | 清华大学 | Text statement processing method and device, computer equipment and storage medium |
CN112528008A (en) * | 2020-12-07 | 2021-03-19 | 北京健康有益科技有限公司 | Diabetic catering method and device based on knowledge graph |
CN112614564A (en) * | 2020-12-24 | 2021-04-06 | 武汉理工大学 | Intelligent nutritional diet recommendation method based on multi-source heterogeneous data fusion |
CN115280422A (en) * | 2020-12-25 | 2022-11-01 | 京东方科技集团股份有限公司 | Food recommendation method and device, storage medium and electronic equipment |
CN112818222A (en) * | 2021-01-26 | 2021-05-18 | 吾征智能技术(北京)有限公司 | Knowledge graph-based personalized diet recommendation method and system |
CN113611393A (en) * | 2021-08-20 | 2021-11-05 | 平安国际智慧城市科技股份有限公司 | Food recommendation, device, equipment and storage medium based on artificial intelligence |
CN114429809A (en) * | 2022-01-25 | 2022-05-03 | 武汉科瓴智能科技有限公司 | Individual nutrition scheme generation method and system for diabetic patients |
CN114758749A (en) * | 2022-03-23 | 2022-07-15 | 清华大学 | Nutritional diet management map creation method and device based on gestation period |
CN114861917A (en) * | 2022-04-29 | 2022-08-05 | 华中科技大学 | Knowledge graph inference model, system and inference method for Bayesian small sample learning |
CN115602289A (en) * | 2022-10-13 | 2023-01-13 | 中南大学(Cn) | Dynamic diet recommendation method based on crowd attribute-diet knowledge graph |
CN115714001A (en) * | 2022-11-14 | 2023-02-24 | 北京航空航天大学 | Method for constructing multi-modal knowledge graph service platform for healthy diet |
CN116127089A (en) * | 2022-12-09 | 2023-05-16 | 海信视像科技股份有限公司 | Intelligent device, menu recommendation method and medium |
CN116166812A (en) * | 2022-12-12 | 2023-05-26 | 中国科学院自动化研究所 | Knowledge graph completion method and device, electronic equipment and storage medium |
Non-Patent Citations (3)
Title |
---|
Identifying Ingredient Substitutions Using a Knowledge Graph of Food;Sola S. Shirai 等;Frontiers in artificial intelligence;第3卷;全文 * |
基于CTD-BLSTM的医疗领域中文命名实体识别模型;祝锡永 等;计算机系统应用(第08期);第173-178页 * |
知识图谱嵌入模型中的损失函数研究综述;申秋慧 等;计算机科学;第50卷(第04期);第149-158页 * |
Also Published As
Publication number | Publication date |
---|---|
CN116417115A (en) | 2023-07-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN116417115B (en) | Personalized nutrition scheme recommendation method and system for gestational diabetes patients | |
CN111415740B (en) | Method and device for processing inquiry information, storage medium and computer equipment | |
Kaur et al. | Deep neural network for food image classification and nutrient identification: A systematic review | |
CN114722208B (en) | Automatic classification and safety level grading method for health medical texts | |
CN117133407B (en) | Nutritional balance assessment method and system for multi-label neural network for children | |
Kaur et al. | Food image-based nutritional management system to overcome polycystic Ovary Syndrome using DeepLearning: A systematic review | |
CN110119479A (en) | A kind of restaurant recommendation method, apparatus, equipment and readable storage medium storing program for executing | |
CN117198466A (en) | Diet management method and system for kidney disease patients | |
Li et al. | Health-guided recipe recommendation over knowledge graphs | |
Li et al. | Deep cooking: Predicting relative food ingredient amounts from images | |
Zhou et al. | CMRDF: A real-time food alerting system based on multimodal data | |
Mejía et al. | Intelligent system for the visual support of caloric intake of food in inhabitants of a smart city using a deep learning model | |
CN112967807A (en) | System, device and storage medium for predicting cerebral apoplexy based on eating behavior | |
US20230255556A1 (en) | Biosensors and food logger systems for personalized health and self-care | |
Al-Saffar et al. | Nutrition information estimation from food photos using machine learning based on multiple datasets | |
Bai et al. | Recommendation algorithm based on probabilistic matrix factorization with adaboost | |
Begum et al. | Artificial Intelligence-Based Food Calories Estimation Methods in Diet Assessment Research | |
Rout et al. | Machine learning model for awareness of diet recommendation | |
JP2022106224A (en) | Program, learning model generation method, and information processing method | |
Chhabra et al. | Artificial Intelligence and Machine Learning Approaches for Understanding Food and Nutrients | |
Najaflou et al. | Recommended system for controlling malnutrition in Iranian children 6 to 12 years old using machine learning algorithms | |
CN112035567A (en) | Data processing method and device and computer readable storage medium | |
US20240087719A1 (en) | Methods and systems for calculating an edible score in a display interface | |
Iheanacho et al. | Classification and recommendation of food intake in West Africa for healthy diet using Deep Learning | |
CN116580817B (en) | Personalized nutrition guidance method and system based on machine learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |